Blog / Your Bidding Algorithm Only Knows What You Tell It

Smart bidding does exactly what it’s asked to do, which can turn into a problem.

“If you are telling Google that you want a higher volume of form fills, demo bookings, downloads, and you have a set budget for that campaign, that’s what Google will do,” says Ashleigh Rowles, senior paid media manager at Somebody Digital. “It will go and optimize towards that form fill.”

What it can’t do is work out which of those form fills was worth anything. “If we don’t apply any value rules or give it any kind of context, it’s going to go and get those leads for you, whether it’s high value, low value, cheap traffic, expensive traffic,” Rowles says. “It doesn’t know that it needs to differentiate.”

Without that signal, the account is running on guesswork. “If we don’t have the correct data being fed back into the accounts, then we are essentially going blind.”

Feeding Offline CRM Conversion Data to Smart Bidding

Rowles keeps the requirements short.

The first is offline conversion data flowing from the CRM back into the platforms, so the algorithm learns what happened to a lead after the form was submitted. The second is a consistent scoring standard for MQLs and SQLs, applied the same way every time, because inconsistent grading feeds the algorithm noise dressed up as signal.

Both are unglamorous plumbing for B2B marketers, unfortunately, neither is optional if you want bidding to chase quality.

Feeding in disqualified leads matters as much as feeding in good ones, since a profile of who not to target sharpens targeting from both directions.

Using Proxy Metrics for B2B Smart Bidding in Low-Volume Accounts

There’s an obvious catch in B2B, and John Wilkes put it to Rowles directly: optimize deeper and you often end up with too few conversions for the algorithm to learn from.

Her answer is refreshingly free of spin. “Short answer, no,” she says, on whether smart bidding copes with low volume. “If you have a very niche product or a low budget, and you’re getting only 3 MQLs a week, it’s not enough to have a really well-oiled smart bidding campaign.”

So in those accounts the campaign keeps optimizing towards lead volume, with a proxy built from CRM ratios sitting behind it. If you know roughly 20% of leads become MQLs, you work back from there to a target that approximates quality. Rowles is upfront about the weakness. “It can be wishy-washy, things can change. It’s something we would have to review every quarter. It’s obviously not ideal.”

An imperfect proxy reviewed quarterly still beats a volume target that hasn’t been questioned in 2 years.